In the febrile markets of early 2026, few documents have stirred as much unease as the Citrini Research memo released this week.
Framed as a “thought exercise in financial history from the future,” the report by analysts Alap Shah and his team painted a chilling hypothetical of a 2028 “Global Intelligence Crisis.”
At its core, the report envisions artificial intelligence’s unchecked ascent triggering mass white-collar displacement and economic stasis. In the scenario, unemployment surges to 10.2 per cent in the US, as AI agents supplant coders, analysts, and intermediaries.
If AI cannot generate broader societal gains alongside productivity gains, it doesn’t justify the label “intelligence” at all — it becomes self-cannibalising capital. A parasitical “anti-intelligence” system.
The S&P 500 plummets 38 per cent from its 2026 highs, erasing trillions in value and echoing the depths of the Great Financial Crisis. Markets, initially buoyed by AI euphoria — pushing the index near 8,000 and the Nasdaq above 30,000 — crumble under the weight of reality: a 2 per cent daily drop on grim job data, an 18 per cent tumble for software giants like ServiceNow, and a 9 per cent hit to payments firms like Mastercard as AI erodes fee-based moats.
The projection is a self-reinforcing “human intelligence displacement spiral.” AI boosts productivity to 1950s levels, generating output that inflates nominal growth but evaporates in the real economy.
Machines, after all, do not dine out, book holidays, or splurge on gadgets; they produce without consuming, flatlining money velocity and gutting the 70 per cent of GDP driven by consumer spending.
Displaced professionals — responsible for three-quarters of discretionary outlays — flood into gig and service jobs, compressing wages across the board. Real wage growth collapses, even as gains accrue to capital owners and compute barons. Globally, the shockwaves are brutal: India’s $200bn IT export engine, a linchpin of its current account surplus, vaporises as AI coding agents slash costs to mere electricity bills. Contracts cancel en masse, the rupee sheds 18 per cent against the dollar in months, and New Delhi scrambles for IMF aid by early 2028.
The scenario was scary enough to jolt markets out of their AI complacency. In U.S. trading, the S&P 500 slid around 1 percent, the Nasdaq 100 fell roughly 1.2 percent, and the Dow Jones dipped about 1.7 percent as investors reassessed risk appetite amid AI disruption narratives. Software and technology-linked shares — including DoorDash and IBM — were among the hardest hit, and defensive assets such as Treasuries and gold saw inflows, with 10-year yields edging modestly lower. Analysts and journalists were quick to link the weakness to heightened anxiety about potential economic effects of automation rather than firm-specific news.
Except there’s a fundamental stupidity at the heart of the report (which, by the way, looks very much like it was written with AI-assistance). I’d argue that what Citrini was really describing was a fundamental paradox at the heart of AI, better thought of as the rise of “anti-intelligence.”
This is a scenario in which we mistake one dimension of intelligence for the whole. Instead of a perfectly weighted “positive sum” balance, the system becomes saturated with what Cambridge psychologist Simon Baron-Cohen calls “systemising” intelligence — rule-based, analytical, mechanically precise — while at the same time progressively eroding its counterweight: “empathising intelligence”.
But real intelligence, whether in a person or an economy, is not maximised by overclocking one faculty at the expense of the other. When systemising capacity expands without sufficient empathic judgment, optimisation becomes brittle. Outcomes may appear efficient in spreadsheet terms, yet prove socially destabilising and economically self-defeating. A system that cannot balance both does not become smarter. It becomes dangerously lopsided.
If AI deployment results in weaker demand, lower aggregate incomes, and a hollowed-out consumer base by over-focusing on systemising intelligence, the outcome does not lead to an efficiency revolution. Rather, it leads to disaster just as envisioned by Citrini.
Fundamentally, if AI cannot generate broader societal gains alongside productivity gains, it doesn’t justify the label “intelligence” at all — it becomes self-cannibalising capital. A parasitical “anti-intelligence” system.
Indeed, I would put it even more bluntly. Any innovation that erodes the very market on which it depends cannot, by definition, be a rational long-term investment.
Dumb finance in the AI age
What Monday’s market tremor actually exposed was a deeper structural fragility in modern finance — one whose negative-sum social consequences remain widely underappreciated. Chief among them is the system’s over-dependence on high-frequency trading algorithms that increasingly set marginal prices on fundamentally mindless terms.
Unlike long-term capital allocators, these systems are optimised for spread capture, volatility harvesting and basis-point extraction — not for assessing whether the underlying economy remains solvent and demand-generative. Market-making and high-frequency strategies exemplify the dynamic: relentless arbitrage and liquidity provision divorced from any stake in long-term value creation. They monetise micro-inefficiencies while remaining indifferent to macro-fragility.
Yet large-scale labour displacement is not a rounding error — it is the ultimate economic inefficiency. An economy cannot compound if the very consumers who sustain revenues are systematically weakened. Price discovery that ignores this reality is not sophisticated; it is structurally stupid.
The result is the opposite of savvy investing. At best, it represents a form of extraction masquerading as efficiency. At worst, it is a form of algorithmic self-sabotage — a system that inflates zero-sum asset bubbles while neglecting the fundamentals that ultimately anchor valuation.
Remember, if the underlying income base erodes, financial engineering simply pulls forward returns from a shrinking future. This is the opposite of value-adding investment.
A new social contract
The answer, however, is not to halt AI adoption. It is to price its externalities properly. The easiest solution is for governments to act swiftly and mandate that AI-driven redundancies carry a structural penalty — not in the form of punitive taxation, but in the form of equity-based compensation for those technologically displaced.
If a role disappears because an algorithm can now perform it, the company should be required to compensate that worker with dividend-yielding perpetual equity calibrated to reflect the productivity gains created by that displacement.
That is to say, if labour is being replaced by capital, then labour must receive a share of that capital.
This is not welfare, nor is it a form of taxpayer-funded Universal Basic Income that saps human agency by reducing everyone to a common denominator. Rather, it is a corporate capital-allocation discipline. Firms already pay redundancy in cash. But cash is a short-term salve. Equity aligns incentives and acknowledges a deeper truth: the AI system only works because it stands on decades of accumulated human knowledge and training data generated by the very professionals it now renders obsolete. That is a form of intellectual inheritance, and inheritance warrants return.
Under such an equity-compensation regime, companies would face a clear choice. They could retain technologically redundant employees on payroll, using enhanced productivity to fund salaries while redeploying them into knowledge preservation, training, governance, or simply labour hoarding until expansion justifies natural attrition. Or they could remove them — but at the cost of equity dilution proportional to the productivity gain achieved.
The result would be rational restraint. Firms would not fire simply to “do the same with less.” They would only displace workers when the incremental return on capital genuinely exceeded the cost of dilution. AI would therefore incentivise expansion and new markets rather than contraction and demand destruction.
Critically, this rule would apply only to genuine technological obsolescence. If an employee is incompetent, redundant due to non-AI-related restructuring, or replaced by another human, standard labour law suffices. But if the job itself vanishes because of AI, aka a “labour for capital substitution” event, then the labour component should be capitalised to equal effect.
For the individual, the choice remains free. They may hold the equity and enjoy leisure supported by dividend flows; sell it to fund retraining or entrepreneurial ventures; or reinvest elsewhere. Agency is preserved. The transition is funded not by the state, but by the productivity windfall that made the displacement possible, without any destabilising social fallout.
Universal Basic Income, by contrast, would socialise the cost while flattening merit. Yet even in a highly automated system, merit must still be incentivised to preserve social cohesion, stability and peace. A model that dulls those incentives risks setting society backwards — eroding, rather than strengthening, the capital and resources on which prosperity depends.
It fundamentally disregards the 20 or 40 years of human capital accumulated by professionals whose expertise underpins and, in many cases, trained the very systems now replacing them. It frames technological displacement as an unfortunate by-product of progress rather than what it truly is: a monetisable transfer of value from labour to capital. Those professionals’ contributions to the system must be acknowledged and conserved.
Of course, if capital is steadily replacing labour as the primary source of returns, then every child should begin life with a baseline capital stake. The recent push toward so-called “Trump accounts” gestures in this direction. But the principle should be more broadly deployed by other governments.
Such endowments would not be welfare. They would be ownership. Each citizen would have the freedom to deploy that stake — for education, enterprise, housing or investment — and bear the consequences of wise or unwise decisions. Agency is preserved; dependency is not institutionalised.
In a world where certain goods remain physically scarce — land, beauty spots, heritage sites, finite resources — some mechanism of disciplined allocation must be maintained, even in an abundant AI future. Broad capital ownership ensures that access to prosperity is not determined purely by inheritance.
More pertinently, the lessons imparted by Henry Ford must be remembered: Mass production only worked because workers were paid enough to buy what they produced. An economy cannot sustainably replace consumers with machines and expect equity valuations to hold.
That is precisely why mandating equity-based penalties for AI-driven redundancy is not the equivalent of Luddism. It is a structural safeguard — one that preserves the integrity of the economic system while still permitting the deployment of productivity-enhancing technology. Properly designed, it ensures that efficiency gains compound into broader social prosperity, abundance, and durable wealth, rather than eroding the very demand base on which that wealth depends.
It is creative destruction without the destruction, because we have reached a stage in the technological cycle where productivity gains are so scalable that expansion no longer requires erasing the income base that sustains demand.
Even Joseph Schumpeter ultimately acknowledged that capitalism’s dynamism depends on the durability of its social foundations. In Capitalism, Socialism and Democracy, he warned that if the social contract frays, the very process of creative destruction can become politically unsustainable — not because it ceases to generate output, but because society ceases to tolerate its consequences.
Fail to embed that logic into policy, and we risk discovering too late that replacing labour with capital at scale does not create abundance — it cannibalises the very market on which capital depends.
Remember: AI that does not serve humanity, but instead requires humanity to serve it, is not AI worth having.
Politicians must act fast.
